Using specific ployclonal antibodies against saxitoxin (STX) or neosaxitoxin (neo-STX) in combination with either STX-horseradish peroxidase (HRP) or neo-STX-HRP, the efficacy of four different direct competitive enzyme-linked immunoassorbent assay (dc-ELISA) formats for the analysis of STX and neo-STX was evaluated. Concentrations causing 50% inhibition (ID 50 ) of binding of toxin-HRP conjugate to the antibodies by free toxins in various ELISAs were found in the range of 0.1-9.0 ng/mL. A dc-ELISA, using either anti-STX/STX-HRP or anti-neo-STX/neo-STX-HRP pairs (ID 50 values of 0.28 and 0.18 ng/mL for STX and neo-STX, respectively), was found to be most effective for the analysis of STX and neo-STX in naturally contaminated shellfish samples. The analytical recoveries of STX added to viscera extracts of butter clams, dungeness crab, tanner crab, and blue mussels in the range of 0.5-10 ng/mL -1 g -1 were found to be 88.1, 92.7, 92.2, and 93.5% with coefficients of variation of 3.9, 2.7, 9.7, and 2.8%, respectively. The detection limit for STX and neo-STX in these shellfish was around 0.2 ng/g of tissue. Gonyautoxins 1-4, but not the C group of PSP toxins, were also detectable in these two systems. Analysis of 154 naturally contaminated shellfish samples showed good correlation between the ELISA (STX plus neo-STX levels) and mouse assay data. The data reported here suggest that simultaneous analysis of both STX and neo-STX by ELISA is necessary for accurate determination of overall PSP toxin levels.